golang pprof性能分析
1. 概述
pprof是go语言自带的性能监控和分析工具。
2. 使用方法
2.1 默认使用方法
package main
import (
"fmt"
"net/http"
_ "net/http/pprof"
"time"
)
func main() {
go func() {
_ = http.ListenAndServe(":6060", nil)
}()
var timer *time.Timer
for {
select {
case <-time.Tick(1 * time.Second):
if timer != nil {
timer.Stop()
timer = nil
}
timer = time.AfterFunc(2*time.Second, timeout)
}
}
}
func timeout() {
fmt.Println("timeout")
}
2.2 自定义mux方法
package main
import (
"fmt"
"github.com/gorilla/mux"
"math"
"net/http"
)
import "net/http/pprof"
func AttachProfiler(router *mux.Router) {
router.HandleFunc("/debug/pprof/", pprof.Index)
router.HandleFunc("/debug/pprof/cmdline", pprof.Cmdline)
router.HandleFunc("/debug/pprof/profile", pprof.Profile)
router.HandleFunc("/debug/pprof/symbol", pprof.Symbol)
}
func SayHello(w http.ResponseWriter, r *http.Request) {
for i := 0; i < 1000000; i++ {
math.Pow(36, 89)
}
fmt.Fprint(w, "Hello!")
}
func main() {
r := mux.NewRouter()
AttachProfiler(r)
r.HandleFunc("/hello", SayHello)
http.ListenAndServe(":6060", r)
}
3. URL说明
| 参数 | PATH | EN | CN |
|---|---|---|---|
| allocs | /debug/pprof/allocs?debug=1 |
A sampling of all past memory allocations | 过去内存分配情况的采样信息 |
| block | /debug/pprof/block?debug=1 |
Stack traces that led to blocking on synchronization primitives | 阻塞操作情况的采样信息 |
| cmdline | /debug/pprof/cmdline |
The command line invocation of the current program | 显示程序启动命令及参数 |
| goroutine | /debug/pprof/goroutine?debug=1 |
Stack traces of all current goroutines | 当前所有协程的堆栈信息 |
| heap | /debug/pprof/heap?debug=1 |
A sampling of memory allocations of live objects. | 活动对象的内存分配的样本 |
| mutex | /debug/pprof/mutex?debug=1 |
Stack traces of holders of contended mutexes | 互斥锁情况的采样信息 |
| profile | /debug/pprof/profile |
CPU profile. | CPU 占用情况的采样信息 |
| threadcreate | /debug/pprof/threadcreate?debug=1 |
Stack traces that led to the creation of new OS threads | 系统线程创建情况的采样信息 |
| trace | /debug/pprof/trace |
A trace of execution of the current program. | 程序运行跟踪信息 |
4. 截图
转载请注明来源，欢迎对文章中的引用来源进行考证，欢迎指出任何有错误或不够清晰的表达。可以在下面评论区评论，也可以邮件至 wind.kaisa@gmail.com
